PREDICTING CUSTOMER ACQUISITION & RETENTION
Sponsored by SiriusXM Radio Inc.
WCAI Research Opportunity
10 February 2012
http://wcai.wharton.upenn.edu @WhartonCAI
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 1
INTRODUCTIONS
David Frear
Executive Vice President and CFO
Joe Zarella
Elea McDonnell Feit
Research Director, WCAI
Eric T. Bradlow
Group Vice President & Chief Service Officer
K.P. Chao Professor
Professor of Marketing, Statistics, and Education
Vice-Dean and Director, Wharton Doctoral Programs
Evelyn Sasmor
Co-Director, WCAI
Vice President, Process & Information Management
Peter S. Fader
Frances and Pei-Yuan Chia Professor
Professor of Marketing
Co-Director, WCAI
Charles Pensig
Assistant Research Director, WCAI
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 2
WHY SHOULD I SUBMIT A PROPOSAL?
• Access to unique data
• Simple, efficient proposal process
• Assistance with data cleaning and preparation
• Teleconference Q&A sessions with the corporate partners
• Invitation to a private symposium to share your research findings with the
corporate partner
• Promotion of your research paper through the WCAI SSRN Research Paper
Series
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 3
S IRIUS XM
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 4
ABOUT S I R I U S XM
Sirius XM Radio is America’s satellite radio company. SiriusXM broadcasts more
than 135 satellite radio channels of commercial-free music, and premier sports,
live news, talk, comedy, entertainment, traffic and weather to more than 21 million
subscribers. SiriusXM offers an array of content from the biggest names in
entertainment, as well as from professional sports leagues, major colleges, and
national news and talk providers.
SiriusXM programming is available on more than 800 devices, including aftermarket and factory-installed radios in cars, trucks, boats, and aircraft, and
consumer electronics products for homes and offices. SiriusXM programming is
also available online at siriusxm.com, and through the SiriusXM Internet Radio
App for Apple, BlackBerry, and Android-powered smartphones and mobile devices.
SiriusXM has arrangements with every major automaker and its radio products are
available for sale at shop.siriusxm.com as well as retail locations nationwide.
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 5
ABOUT S I R I U S XM MARKE TI NG AND ANALYTI C S
Mission
• Maximize subscribers and revenue
Organizational Goals
•
•
•
•
•
•
Overcome people’s unwillingness to pay for radio - communicate why it’s worth it
Improve the customer experience
Increase conversions from trial subscriptions to self-pay subscriptions
Reduce subscriber churn
Improve revenue per customer and reduce marketing costs
Communicate with the customers in a way that creates and sustains a relationship that
shows that we know the customer throughout the customer’s lifecycle with us
o
Present the right offer to the right customers at the right time – with “offer” being more
than just price variation
o
Know our customers – get a 360-degree view of the customer’s relationship with us
o
Gather and utilize enhanced customer knowledge to anticipate points of transition and
maintain customer history
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 6
SIRIUSXM SUBSCRIPTION PROCESS: CONVERSION
This is a hypothetical user’s path through subscription. All events are tracked by
SiriusXM.
• When a customer buys a car he either a) already has an active radio in the car
and the sale info is sent to SiriusXM by the car manufacturer, b) calls in to start the
trial, or c) the car dealer activates the trial for the customer. Whichever of a, b, or c
occurs first marks the beginning of the conversion stream
• At the end of the stream of events below, if the customer converts, he enters an
automatically renewing paid-for subscription plan
POST-ACTIVATION
PRE–DEACTIVATION
ENGAGEMENT
END OF TRIAL
RADIO
DEACTIVATES
WELCOME
E-mail
WELCOME
KIT
“Remit”
WELCOME & Engagement Emails
CONVERSION
1&2
OTM*
Approx 35 & 46 days
after trial starts.
E-NEWSLETTERS
CONVERSION CONVERSION
DM
E-MAIL 1
CONVERSION
DM ^
“Remit”
CONVERSION
E-MAIL 2
CONVERSION
DM +
WHARTON CUSTOMER ANALYTICS INITIATIVE
CONVERSION CONVERSION
E-MAIL 3
OTM**
PAGE 7
S I R I U S X M S U B S C R I P T I O N P R O C E S S : P O S T- D E A C T I VATI O N
• If the customer does not convert after the trial period ends, the radio is deactivated
and a post-deactivation campaign commences
• If the customer converts during the post-deactivation campaign, he would enter
an automatically renewing paid-for subscription plan just as if he had converted
during the trial period
• If the customer does not convert during this campaign he becomes part of a
general marketing campaign for reactivations
POST–DEACTIVATION
END OF TRIAL
RADIO
DEACTIVATED
WINBACK
E-MAIL 1
WINBACK
OTM 1^
WINBACK
DM
WINBACK
E-MAIL 2
WINBACK
DM
WINBACK
E-MAIL 3
WINBACK
DM
WINBACK
DM* & OTM**
WHARTON CUSTOMER ANALYTICS INITIATIVE
WINBACK
OTM 2
WINBACK
DM
WINBACK
E-MAIL 4
PAGE 8
S I R I U S XM SUBSCRIPTION PROCESS: WIN BACK
• Finally, if a paying subscriber deactivates his paid-for subscription service he
becomes part of a win back campaign
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 9
IMPORTAN T TERMS
• Churn is the event when a revenue-generating subscriber who
pays out-of-pocket (versus revenue-generating trial subscriptions
sponsored by a car manufacturer) decides to terminate service
for a specific device.
The reported average SiriusXM self-pay monthly churn rate for 2010 was
1.9%.
• Conversion (or Trial Conversion) is the event when a subscriber
who is not paying for a subscription out of his own pocket (either
provided free by SiriusXM or paid for by a car manufacturer)
decides to become a paying subscriber and purchases a
subscription after his trial period ends.
The reported SiriusXM conversion rate for 2010 was 46.2%.
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 10
DATA
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 11
NOTE ON THE SAMPLE DATA
You can look at patterns within a group of like
customers; however, the overall patterns in these data
(e.g., churn or conversion rates) should not be taken as
representative of the full SiriusXM customer base.
Records were selected to provide good data for relevant
scenarios, not to mimic the distributions across the full
customer base.
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 12
DATA : 360 DEGREE VIEW OF CUSTOMER
The data consist of information on 300,000 customers who first activated a trial or
paid subscription in September or October of 2009. Their histories run through the
the end of 2011, providing a full 27 months of history. For each customer we have:
• Devices: what type of device(s) the customer owns, where they were purchased,
and how they were activated/deactivated
• Vehicles: type of car (aliased), time and location of purchase
• Subscription histories: price paid and duration of services, types of services (e.g.
radio-only, internet radio, weather service)
• Outbound marketing: direct marketing efforts made to customers and how they
responded (if relevant)
• Customer service records: use of online self help tool, calls to call center,
including free-text notes taken by call center agents
• Billing records: when bills were sent and paid, and whether customers went into
collections for late payments
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 13
DATA : 360 DEGREE VIEW OF CUSTOMER
• About 12% of households in this sample had more than one active device in
January 2011
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 14
P O TE N TI A L RE S E A R C H Q UE S TI O N S : CH U R N
• Using data on subscriber interactions including the free text in call center notes
and email response notes, can we identify users that are likely to deactivate
soon or who are most likely to convert to a paying customer?
• What are the key customer events, such as calls into the call center, falling into
collections, a certain time in the customer life cycle, the content or reasons for the
calls, use or non-use of self-care site, etc. that cause or precede deactivation?
• What actions on the part of SiriusXM forestall deactivation?
• How does going into collections (late payment) affect churn?
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 15
DATA : CHUR N
• Subscribers choose from a variety of terms lengths with different pricing. The top
5 most-popular plans are:
Plan Type
Cost
Annual
$142.45
Monthly
$12.95
Two Year
$271.95
Quarterly
$38.85
Semi Annual $77.70
* Note: these are 2011 prices which are slightly lower than 2012 prices published on the SiriusXM Website
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 16
DATA : CHUR N
• Because subscriptions are for a fixed amount of time (3, 6, 12 months, etc.)
churning happens most around these “round numbers”.
• However, if the subscriber sells the car, they may deactivate subscriptions “offcycle”
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 17
P O TE N TI A L RE S E A R C H Q UE S TI O N S : CONV E R S I O N
• Can user behavior be used to predict which trial-subscription
customers will convert and target trial-subscribers who will be
positively influenced by additional direct marketing such as
outbound calls? Are there identifiable segments that should be
marketed to differently?
• Can the optimal plan and pricing to convert a trial user be
determined in real time?
• How do the attributes of a trial promotion affect conversion? Do
features such as length of trial and service bundling have an
impact?
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 18
DATA : CONVER SION
• The trial period offered to users varies across users, depending on arrangements
between the car manufacturer and SiriusXM. Most trials are 3, 6 or 12 months in
duration.
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 19
DATA : CONVER SION
• As an example, within a particular group, customers who began trials in H2’10 are
contacted mainly by direct mail during the conversion stream.
• Note that e-mail addresses often are not provided by the OEMs; SiriusXM
collects them once the customer contacts SiriusXM or through other
appends.
• After conversion, subscribers in this cohort are contacted primarily by email
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 20
DATA : CONVER SION
• SiriusXM seeks to better understand the relationship between the number of
marketing touches a customer receives and her likelihood to convert.
• Simply looking at the relationship between marketing touches as conversion
suggests that marketing touches do not improve conversion, however, this
ignores the fact that those subscribers most impressed tend to convert sooner
and so receive fewer communications.
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 21
DATA : TRIAL CONVER SION
• One potential explanatory variable for trial conversion is the car make.
Low %
High %
• For the H2’10 cohort there appears to be substantial differences in conversion
rates between makes
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 22
P O TE N TI A L RE S E A R C H Q UE S TI O N S : COMPE TI TI O N
• How does the availability of terrestrial radio (which varies across
geography) affect subscription activation and deactivation, e.g.
users who like opera, but have no access to terrestrial radio
stations that play opera may have lower propensity to unsubscribe
than users who have good terrestrial options?
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 23
DATA : COMPETITI V E OFFERI NGS
• How do competitive terrestrial (AM/FM) radio offerings affect subscriber churn?
• WCAI will provide a list of a radio stations per zip code with the data set
• The graph below shows cumulative churn over 2 years for the September 2009
cohort
More Terrestrial Radio
Less Terrestrial Radio
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 24
P O TE N TI A L RE S E A R C H Q UE S TI O N S
• Using listening data (available for internet users only), can we
identify patterns in listenership that precede subscription or
deactivation? Can this rich information about the internet users be
used to predict subscription deactivations among the non-internet
users (some principled assumptions will be required)?
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 25
DATA : INTERN ET RADIO
SiriusXM has provided “listening logs” for a sample of its internet radio users
• Internet radio subscribers are currently a small portion SiriusXM’s total user base
• Listening logs record which channels a user listened to, on what device (PC,
smartphone, tablet, other music player) and for how long he listened to each
channel.
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 26
DATA : INTERN ET RADIO
• Because of the time period that was selected for this sample of data, the data
show the majority of devices that use SiriusXM Internet Radio are PCs, followed
by iPhones (~13%). The actual usage on smart phones (iPhone and Android) and
tablets has grown rapidly.
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 27
DATA : INTERN ET RADIO
• A “session” is a continuous period of listening to one internet radio channel (e.g.
90s Rock or Mozart)
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 28
Q&A
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 29
Q& A : PROPOSA L PROCESS
• Review the WCAI FAQ on Research Opportunities:
http://www.wharton.upenn.edu/wcai/dump/FAQ.cfm
• Contact wcai-research@wharton.upenn.edu, if you have questions about the data
• Submit a brief proposal to wcai-research@wharton.upenn.edu by Friday, 2/24/2012.
Proposals should be less than 2,000 words and should include:
o
Research team, affiliations and e-mail addresses
•
Please designate a corresponding author
o
Objectives & contribution to the academic literature
o
Proposed methods
o
Rough timeline
o
Potential for managerial insights & impact for ARC and other non-profits
o
PDF format preferred
o
Please put “RO: SiriusXM Proposal” in the subject line
• Proposals will be evaluated by Elea Feit (WCAI), Pete Fader (WCAI), Eric Bradlow
(WCAI), Kartik Hosanagar (Wharton), Foster Provost (NYU), Evelyn Sasmor (SiriusXM),
and Brian Wood (SiriusXM)
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 30
Q &A: O THE R O P P O RT U N I TI E S FOR RES E A R C H E R S
If you registered for this webinar, you will receive regular announcements about:
• Research Opportunities like this one
o
MARS International: April 2012
o
NC Soft: June 2012
• Grant/funding opportunities
• WCAI Conferences
o
Marketing on the Move: Understanding the Impact of Mobile on Consumer
Behavior, Feb 27-28 at The Wharton School (co-sponsored with MSI)
http://www.wharton.upenn.edu/wcai/MobileConference2012.cfm
• Announcements are also available at http://wcai.wharton.upenn.edu
Visit SSRN Research Paper series: http://www.ssrn.com/link/Wharton-CustAnalytics-Initiative-RES.html
WHARTON CUSTOMER ANALYTICS INITIATIVE
PAGE 31